<s>
In	O
computing	O
,	O
a	O
snowflake	B-Language
schema	I-Language
is	O
a	O
logical	B-Application
arrangement	I-Application
of	O
tables	O
in	O
a	O
multidimensional	O
database	O
such	O
that	O
the	O
entity	B-Architecture
relationship	I-Architecture
diagram	I-Architecture
resembles	O
a	O
snowflake	B-Application
shape	O
.	O
</s>
<s>
The	O
snowflake	B-Language
schema	I-Language
is	O
represented	O
by	O
centralized	O
fact	O
tables	O
which	O
are	O
connected	O
to	O
multiple	O
dimensions	O
.	O
</s>
<s>
"	O
Snowflaking	B-Language
"	O
is	O
a	O
method	O
of	O
normalizing	B-Application
the	O
dimension	O
tables	O
in	O
a	O
star	B-Application
schema	I-Application
.	O
</s>
<s>
When	O
it	O
is	O
completely	O
normalized	B-Application
along	O
all	O
the	O
dimension	O
tables	O
,	O
the	O
resultant	O
structure	O
resembles	O
a	O
snowflake	B-Application
with	O
the	O
fact	O
table	O
in	O
the	O
middle	O
.	O
</s>
<s>
The	O
principle	O
behind	O
snowflaking	B-Language
is	O
normalization	B-Application
of	O
the	O
dimension	O
tables	O
by	O
removing	O
low	O
cardinality	O
attributes	O
and	O
forming	O
separate	O
tables	O
.	O
</s>
<s>
The	O
snowflake	B-Language
schema	I-Language
is	O
similar	O
to	O
the	O
star	B-Application
schema	I-Application
.	O
</s>
<s>
However	O
,	O
in	O
the	O
snowflake	B-Language
schema	I-Language
,	O
dimensions	O
are	O
normalized	B-Application
into	O
multiple	O
related	O
tables	O
,	O
whereas	O
the	O
star	B-Application
schema	I-Application
's	O
dimensions	O
are	O
denormalized	O
with	O
each	O
dimension	O
represented	O
by	O
a	O
single	O
table	O
.	O
</s>
<s>
A	O
complex	O
snowflake	B-Application
shape	O
emerges	O
when	O
the	O
dimensions	O
of	O
a	O
snowflake	B-Language
schema	I-Language
are	O
elaborate	O
,	O
having	O
multiple	O
levels	O
of	O
relationships	O
,	O
and	O
the	O
child	O
tables	O
have	O
multiple	O
parent	O
tables	O
(	O
"	O
forks	O
in	O
the	O
road	O
"	O
)	O
.	O
</s>
<s>
Star	O
and	O
snowflake	B-Language
schemas	I-Language
are	O
most	O
commonly	O
found	O
in	O
dimensional	O
data	O
warehouses	O
and	O
data	B-General_Concept
marts	I-General_Concept
where	O
speed	O
of	O
data	O
retrieval	O
is	O
more	O
important	O
than	O
the	O
efficiency	O
of	O
data	O
manipulations	O
.	O
</s>
<s>
As	O
such	O
,	O
the	O
tables	O
in	O
these	O
schemas	O
are	O
not	O
normalized	B-Application
much	O
,	O
and	O
are	O
frequently	O
designed	O
at	O
a	O
level	O
of	O
normalization	B-Application
short	O
of	O
third	O
normal	B-Application
form	I-Application
.	O
</s>
<s>
Normalization	B-Application
splits	O
up	O
data	O
to	O
avoid	O
redundancy	O
(	O
duplication	O
)	O
by	O
moving	O
commonly	O
repeating	O
groups	O
of	O
data	O
into	O
new	O
tables	O
.	O
</s>
<s>
Normalization	B-Application
therefore	O
tends	O
to	O
increase	O
the	O
number	O
of	O
tables	O
that	O
need	O
to	O
be	O
joined	O
in	O
order	O
to	O
perform	O
a	O
given	O
query	O
,	O
but	O
reduces	O
the	O
space	O
required	O
to	O
hold	O
the	O
data	O
and	O
the	O
number	O
of	O
places	O
where	O
it	O
needs	O
to	O
be	O
updated	O
if	O
the	O
data	O
changes	O
.	O
</s>
<s>
This	O
often	O
negates	O
the	O
potential	O
storage-space	O
benefits	O
of	O
the	O
star	B-Application
schema	I-Application
as	O
compared	O
to	O
the	O
snowflake	B-Language
schema	I-Language
.	O
</s>
<s>
Example	O
:	O
One	O
million	O
sales	O
transactions	O
in	O
300	O
shops	O
in	O
220	O
countries	O
would	O
result	O
in	O
1,000,300	O
records	O
in	O
a	O
star	B-Application
schema	I-Application
(	O
1,000,000	O
records	O
in	O
the	O
fact	O
table	O
and	O
300	O
records	O
in	O
the	O
dimensional	O
table	O
where	O
each	O
country	O
would	O
be	O
listed	O
explicitly	O
for	O
each	O
shop	O
in	O
that	O
country	O
)	O
.	O
</s>
<s>
A	O
more	O
normalized	B-Application
snowflake	B-Language
schema	I-Language
with	O
country	O
keys	O
referring	O
to	O
a	O
country	O
table	O
would	O
consist	O
of	O
the	O
same	O
1,000,000	O
record	O
fact	O
table	O
,	O
a	O
300	O
record	O
shop	O
table	O
with	O
references	O
to	O
a	O
country	O
table	O
with	O
220	O
records	O
.	O
</s>
<s>
Some	O
database	O
developers	O
compromise	O
by	O
creating	O
an	O
underlying	O
snowflake	B-Language
schema	I-Language
with	O
views	B-Language
built	O
on	O
top	O
of	O
it	O
that	O
perform	O
many	O
of	O
the	O
necessary	O
joins	O
to	O
simulate	O
a	O
star	B-Application
schema	I-Application
.	O
</s>
<s>
This	O
provides	O
the	O
storage	O
benefits	O
achieved	O
through	O
the	O
normalization	B-Application
of	O
dimensions	O
with	O
the	O
ease	O
of	O
querying	O
that	O
the	O
star	B-Application
schema	I-Application
provides	O
.	O
</s>
<s>
The	O
snowflake	B-Language
schema	I-Language
is	O
in	O
the	O
same	O
family	O
as	O
the	O
star	B-Application
schema	I-Application
logical	O
model	O
.	O
</s>
<s>
In	O
fact	O
,	O
the	O
star	B-Application
schema	I-Application
is	O
considered	O
a	O
special	O
case	O
of	O
the	O
snowflake	B-Language
schema	I-Language
.	O
</s>
<s>
The	O
snowflake	B-Language
schema	I-Language
provides	O
some	O
advantages	O
over	O
the	O
star	B-Application
schema	I-Application
in	O
certain	O
situations	O
,	O
including	O
:	O
</s>
<s>
Some	O
OLAP	B-Application
multidimensional	O
database	O
modeling	O
tools	O
are	O
optimized	O
for	O
snowflake	B-Language
schemas	I-Language
.	O
</s>
<s>
Normalizing	B-Application
attributes	O
results	O
in	O
storage	O
savings	O
,	O
the	O
tradeoff	O
being	O
additional	O
complexity	O
in	O
source	O
query	O
joins	O
.	O
</s>
<s>
The	O
primary	O
disadvantage	O
of	O
the	O
snowflake	B-Language
schema	I-Language
is	O
that	O
the	O
additional	O
levels	O
of	O
attribute	O
normalization	B-Application
adds	O
complexity	O
to	O
source	O
query	O
joins	O
,	O
when	O
compared	O
to	O
the	O
star	B-Application
schema	I-Application
.	O
</s>
<s>
Snowflake	B-Language
schemas	I-Language
,	O
in	O
contrast	O
to	O
flat	O
single	O
table	O
dimensions	O
,	O
have	O
been	O
heavily	O
criticised	O
.	O
</s>
<s>
Their	O
goal	O
is	O
assumed	O
to	O
be	O
an	O
efficient	O
and	O
compact	O
storage	O
of	O
normalised	B-Application
data	I-Application
but	O
this	O
is	O
at	O
the	O
significant	O
cost	O
of	O
poor	O
performance	O
when	O
browsing	O
the	O
joins	O
required	O
in	O
this	O
dimension	O
.	O
</s>
<s>
When	O
compared	O
to	O
a	O
highly	O
normalized	B-Application
transactional	O
schema	O
,	O
the	O
snowflake	B-Language
schema	I-Language
's	O
denormalization	O
removes	O
the	O
data	O
integrity	O
assurances	O
provided	O
by	O
normalized	B-Application
schemas	O
.	O
</s>
<s>
Data	O
loads	O
into	O
the	O
snowflake	B-Language
schema	I-Language
must	O
be	O
highly	O
controlled	O
and	O
managed	O
to	O
avoid	O
update	O
and	O
insert	B-Application
anomalies	I-Application
.	O
</s>
<s>
The	O
example	O
schema	O
shown	O
to	O
the	O
right	O
is	O
a	O
snowflaked	O
version	O
of	O
the	O
star	B-Application
schema	I-Application
example	O
provided	O
in	O
the	O
star	B-Application
schema	I-Application
article	O
.	O
</s>
<s>
The	O
following	O
example	O
query	O
is	O
the	O
snowflake	B-Language
schema	I-Language
equivalent	O
of	O
the	O
star	B-Application
schema	I-Application
example	O
code	O
which	O
returns	O
the	O
total	O
number	O
of	O
television	O
units	O
sold	O
by	O
brand	O
and	O
by	O
country	O
for	O
1997	O
.	O
</s>
<s>
Notice	O
that	O
the	O
snowflake	B-Language
schema	I-Language
query	O
requires	O
many	O
more	O
joins	O
than	O
the	O
star	B-Application
schema	I-Application
version	O
in	O
order	O
to	O
fulfill	O
even	O
a	O
simple	O
query	O
.	O
</s>
<s>
The	O
benefit	O
of	O
using	O
the	O
snowflake	B-Language
schema	I-Language
in	O
this	O
example	O
is	O
that	O
the	O
storage	O
requirements	O
are	O
lower	O
since	O
the	O
snowflake	B-Language
schema	I-Language
eliminates	O
many	O
duplicate	O
values	O
from	O
the	O
dimensions	O
themselves	O
.	O
</s>
